首页 | 本学科首页   官方微博 | 高级检索  
     


An efficient approach for the multiprocessor non-preemptive strictly periodic task scheduling problem
Affiliation:1. School of Microelectronics, Shanghai Jiao Tong University, 800 DongChuan Road, Shanghai, China;2. Department of Computer Science, Pace University, 1 Pace Plaza, Manhattan, New York City, NY 10038, USA;1. Eindhoven University of Technology and Centrum Wiskunde & Informatica, Netherlands;2. AUDI AG, Germany;3. Chalmers | University of Gothenburg, Sweden
Abstract:Strict periodicity constraint is of great importance since it concerns some hard real-time systems where missing deadlines leads to catastrophic situations. However, the problem of schedulability analysis for non-preemptive strictly periodic tasks on a multiprocessor platform is even more intractable than the one with the common periodicity. In order to implement such systems, designers need effective tools based on fast and near-optimal solutions.This paper presents a schedulability analysis which results mainly in a, two versions, task assignment and start-time calculation algorithm. The first one targets the harmonic task periods case while the second one targets the non-harmonic task periods case. Each version is based on a sufficient uniprocessor schedulability test. In addition, for the non-harmonic case which is the most intractable, the uniprocessor sufficient schedulability test uses the strictly periodic task utilization factor. This factor stands for the fraction of time spent to execute a task while its strict periodicity and the ones of the already scheduled tasks are met. As a result, an efficient and easily implementable scheduling algorithm is proposed which begins by assigning tasks to processors then attributes a start-time to every task in such a way that strict periodicity and deadline constraints are met. The effectiveness of the proposed scheduling algorithm, in both versions, has been shown by a performance evaluation and comparisons with an optimal and a similar suboptimal solution.
Keywords:
本文献已被 ScienceDirect 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号